Google inadvertently released a blog post on June 25th announcing Gemini CLI, a new open-source project. The post detailed the tool’s capabilities before it was swiftly taken down, replaced by a 404 error. Developer Meet Patel on X (formerly Twitter) quickly captured a screenshot and provided an archive.org link, preserving the information before it was removed. The blog post revealed Google’s plan to integrate its Gemini AI models directly into developer terminals through a free command-line tool. Gemini CLI offers the Gemini 2.5 Pro model with a 1 million token context. Key features include 60 model requests per minute and 1,000 per day for free with a Google login, integration with Gemini Code Assist for VS Code users, real-time web grounding via Google Search, and support for extensions, scripting, automation, and system prompts. This tool is designed to streamline developer workflows by reducing the need to switch between various tools. It can generate code, debug, and facilitate experimentation, as well as handle research tasks and generate short videos. As an open-source project under Apache 2.0, developers can audit the code, create extensions, and customize workflows. The reason for the post’s removal remains unclear, although the fact that it was authored by senior Google engineers Taylor Mullen and Ryan Salva suggests it was not a fabrication. The cached post still offers insight into what developers can expect from AI agents.
